Using the National Weather Service's (NWS) tornado verification data set (warnings and events) which cover a period of 19 years from 1986 to 2004, NWS statistics show that mean tornado warning lead time (amount of time between warning issuance and tornado event) has increased from around 5 minutes to about 13 minutes. This result depends on how one deals with tornadoes that occur before a warning is issued. The NWS counts those as having a lead time of zero minutes. By using that same data set and considering only those events with positive lead time, the lead time has stayed roughly the same at around 18.5 minutes, with the probability of detection (POD) increasing from approximately 25% to 75%. The apparent increase in the NWS lead time numbers is a result of multiplying the increasing POD with the constant 18.5 minute lead time. The mean lead time for warned tornadoes is also constant for different F-scale ratings, with the higher F-scale rated events having higher POD.
We have also looked at the duration of tornado warnings, and as it turns out the average person in the US spends about .6 hours or approximately 36 minutes, a year, under tornado warnings. This brings into question the importance of false alarms in evaluating the warning system. The maximum annual average time under warning for any county in the US was 5.8 hours, in Harris County, Texas, which includes the city of Houston. Combining this with the large population (over 3.6 million, the third most populous county in the country), Harris County has the maximum number of “warned people–hours” by far with almost 15 million per year, or about 8% of the national total.